Biological Interpretation System

After filtering, ranking, identifier cleaning, enrichment preparation, STRING input preparation, and visualization, the next step is biological interpretation.

This chapter brings the outputs together into a structured interpretation system.

The goal is not only to list significant proteins. The goal is to explain what the proteomics results suggest biologically, while clearly separating evidence from interpretation.

Starting Point

This chapter uses outputs from earlier workflow steps:

results/differential-summary.tsv
results/ranking-summary.tsv
results/identifier-cleaning-summary.tsv
results/enrichment-input-summary.tsv
results/string-network-summary.tsv
results/top-significant-proteins-for-plot.tsv
results/ranked-significant-proteins.tsv

It also refers to figures created in Chapter 10:

results/figures/volcano-plot.png
results/figures/top-significant-proteins.png
results/figures/regulation-summary.png

Why an Interpretation System Is Needed

Proteomics analysis can generate many files.

Without a structured interpretation system, the final result can become a collection of tables and plots without a clear biological message.

A good interpretation system asks:

What changed?
        ↓
How strong was the change?
        ↓
Which proteins changed most?
        ↓
Which biological functions are suggested?
        ↓
Are proteins connected in networks?
        ↓
Does the result support the study question?
        ↓
What are the limitations?

This chapter provides a reproducible structure for answering those questions.

Evidence Layers

The biological interpretation should be built from multiple layers of evidence.

Differential abundance results
        ↓
Ranked proteins
        ↓
Identifier cleaning
        ↓
Functional enrichment inputs
        ↓
STRING network inputs
        ↓
Visualization
        ↓
Interpretation summary

Each layer contributes something different.

Layer 1: Differential Abundance Summary

The differential abundance summary answers:

How many proteins were tested?
How many passed the significance thresholds?
How many were upregulated?
How many were downregulated?
What thresholds were used?

The relevant file is:

results/differential-summary.tsv

This is the first checkpoint for interpretation.

Layer 2: Ranked Protein Review

The ranked protein tables answer:

Which proteins should be reviewed first?
Which proteins had the strongest statistical support?
Which proteins had the largest abundance changes?
Are top proteins biologically plausible?

The key file is:

results/ranked-significant-proteins.tsv

A smaller plot-ready table is:

results/top-significant-proteins-for-plot.tsv

Layer 3: Identifier and Annotation Readiness

The identifier-cleaning summary answers:

How many identifiers were usable?
How many gene symbols were available?
Were there possible contaminants?
Were there reverse or decoy entries?
Were protein groups present?

The relevant file is:

results/identifier-cleaning-summary.tsv

Interpretation should be cautious when many identifiers are missing or ambiguous.

Layer 4: Functional Enrichment Readiness

The enrichment input summary answers:

How many identifiers are available for enrichment?
Are there enough upregulated proteins?
Are there enough downregulated proteins?
Should enrichment be run separately by direction?

The relevant file is:

results/enrichment-input-summary.tsv

This chapter does not claim enrichment results unless enrichment has actually been run.

Instead, it documents whether the dataset is ready for enrichment analysis.

Layer 5: STRING Network Readiness

The STRING summary answers:

How many identifiers are ready for STRING?
Are there separate upregulated and downregulated lists?
Is the network input large enough to be informative?

The relevant file is:

results/string-network-summary.tsv

STRING interpretation should only be written after network analysis has been performed and exported.

Layer 6: Visual Summary

The visual summary helps communicate the results.

Useful figures include:

volcano-plot.png
top-significant-proteins.png
regulation-summary.png

These figures support interpretation but should not replace careful biological reasoning.

Interpretation Script

The executable script is saved as:

scripts/R/11-build-biological-interpretation-summary.R

The script gathers available workflow summaries and creates a draft interpretation summary.

The guide shows how to run the script. The full executable code is maintained in scripts/R/.

Running the Script

From the project root, run:

Rscript scripts/R/11-build-biological-interpretation-summary.R \
  results \
  results/biological-interpretation-summary.md

The arguments are:

1st argument → results directory
2nd argument → output markdown summary

Output File

The script creates:

results/biological-interpretation-summary.md

This file is a structured markdown summary that can be reviewed, edited, and included in the final report.

Interpretation Summary Structure

The generated summary includes sections for:

Analysis overview
Differential abundance summary
Top-ranked proteins
Identifier readiness
Functional enrichment readiness
STRING network readiness
Available figures
Interpretation notes
Limitations
Next steps

This creates a reusable bridge between computational outputs and final reporting.

Evidence-Based Interpretation

A useful biological interpretation should distinguish between:

observed results
        vs
biological interpretation
        vs
hypothesis or mechanism

For example:

Observed result:
Several proteins show higher abundance in treated samples.

Biological interpretation:
The top proteins may suggest altered mitochondrial or metabolic activity.

Hypothesis:
The treatment may affect energy metabolism.

Only the first statement is directly observed from the table. The later statements require supporting evidence.

Avoiding Overinterpretation

Proteomics results should not be overclaimed.

Avoid statements like:

This proves that the pathway is activated.

Prefer cautious statements:

The differentially abundant proteins suggest involvement of this pathway.

or:

These results are consistent with changes in this biological process.

Manual Interpretation Questions

After reviewing the generated summary, ask:

Are the top proteins biologically coherent?
Do upregulated and downregulated proteins suggest different themes?
Are the findings consistent with the study design?
Are there possible technical explanations?
Are contaminants or protein groups affecting interpretation?
Do enrichment and STRING results support the same biological story?

These questions help convert results into a defensible interpretation.

Recommended Interpretation Language

Useful wording includes:

The results suggest...
The protein list is consistent with...
The top-ranked proteins include...
The enrichment-ready inputs indicate...
The STRING-ready identifiers support follow-up network analysis...
The interpretation should be treated cautiously because...

This style is appropriate for early reports, exploratory analyses, and client-facing summaries.

Limitations to Record

A transparent interpretation should record limitations such as:

small sample size
missing identifiers
limited annotation coverage
protein groups instead of single proteins
lack of raw-data reprocessing
unknown batch effects
threshold-dependent protein selection
lack of validation experiment

Limitations improve trust in the report.

Relationship to the Final Report

The output from this chapter becomes part of the final reproducible report.

biological-interpretation-summary.md
        ↓
reproducible proteomics report

The final report can include:

methods summary
tables
figures
interpretation notes
limitations
next steps

Looking Ahead

The next chapter creates a reproducible proteomics report.

Structured interpretation
        ↓
Report-ready summary
        ↓
Reproducible proteomics report
        ↓
Reusable analysis system